Transition-Aware best-of-N sampling for Longitudinal Chest X-ray Reports

2026-06-23

AI Summary

This work addresses the limitation of existing chest X-ray report generation methods, which overlook semantic changes between current and prior examinations in longitudinal clinical settings. The authors propose a training-free best-of-N sampling strategy that, for the first time, incorporates a longitudinal change-aware mechanism into report generation. By embedding report sentences as sets and combining four directional set distances—mean shift, novelty residual, directed Hausdorff anchor, and cost-weighted optimal transport—with cosine similarity, the method ranks candidate reports using a library of real transition vectors. Evaluated on a multi-visit AP-PA dataset, the approach consistently outperforms random selection across three prompting configurations and three vision-language models, with the most pronounced improvements observed in the impression section, all without fine-tuning the underlying pretrained models.

Abstract

In longitudinal clinical practice, every chest X-ray is read in the context of the patients prior exam, and much of what the radiologist communicates is the change from one visit to the next. To the best of our knowledge, we present the first training-free best-of-N sampling scheme for pre-trained chest X-ray report generators that is explicitly aware of this longitudinal prior to current transition. We call it transition-aware best-of-N sampling, each report is split into sentences and embedded into an unordered set in Rd; each (prior, current) pair is reduced to a fixed-dim directional vector via a set-to-set distance designed to encode the change between the two sets; and candidates are scored by cosine distance from their candidate transition vector to a cached bank of ground-truth training transition vectors, aggregated as min or kNN. We instantiate the framework with four directional set distances (mean-shift, novelty residual, directed-Hausdorff anchor, and cost-weighted optimal transport) and evaluate on a multi-visit AP-PA cohort, running inference under three prompts on three vision-language generators. Transition-aware best-of-N outperforms random selection across the board, with the largest relative gains on the Impression section.

Problem

• longitudinal
• chest X-ray reports
• transition-aware
• report generation
• temporal change

Innovation

• transition-aware sampling
• longitudinal chest X-ray
• set-to-set distance
• best-of-N decoding
• training-free adaptation